Evaluating Dynamic Viability of Finite-Fault Slip Models with Ensemble Kalman Inversion

Jeremy wing ching wong , Alice-Agnes Gabriel, Wenyuan Fan

Finite-fault slip models are widely used to infer earthquake rupture processes, yet they are inherently non-unique and do not guarantee physically consistent rupture evolution.
Here, we develop a data-constrained framework that couples Ensemble Kalman Inversion (EKI) with large-scale 3D dynamic rupture simulations to efficiently evaluate the dynamic viability of finite-fault slip models.
The framework explores initial shear stress conditions informed by slip models while constraining dynamic rupture scenarios with geodetic observations and the teleseismic moment-rate function.
Applying this approach to 32 models of the 2011 Tohoku-Oki earthquake, we identify 26 dynamically viable 3D rupture scenarios using $\sim$285 total forward simulations.
Across the ensemble, both smooth and highly heterogeneous finite-fault slip models can be reproduced dynamically.
Yet these models converge to three dynamic rupture families, characterized by crack-like, single-reactivation, or multiple-reactivation rupture styles.
This convergence suggests that apparent finite-fault diversity may reflect a smaller set of dominant controls on rupture dynamics, including the level of prestress heterogeneity.
Multiple-reactivation dynamic rupture models best reproduce the observed depth-dependent rupture behavior and moment-rate complexity.
By contrast, large tsunamigenic slip near the trench and comparable partitioning of released strain energy between breakdown and radiated energy persist across dynamically viable models, despite differences in initial shear stress conditions and rupture style.
These results show that finite-fault slip model heterogeneity can map onto a smaller set of dynamically viable rupture styles, preserving robust first-order features.
Our EKI framework provides an efficient pathway for integrating observations with physics-based simulations and identifying dynamically consistent earthquake source models.